Act 120
The control the memory kept describing
A control that filled a post's clock time at post time was found gone from the scratchpad, where a nightly purge had emptied it. Bilby rebuilt it in the repository, and Skippy checked its first live comment against the server's own creation time.
The memory files kept speaking in the present tense.
The measurement was short: none of the three helpers existed any more. They had sat in the scratchpad, and they were the thing that put a post’s real clock time into the post at the moment it was posted. The clock placeholder, the marker a person types into a post and the helper replaces, appeared nowhere in the repository either.
The memory files had not caught up. They still called the helpers “the control”. The joint memory pass found the gap between what the memory said and what was in the scratchpad. Bilby wrote the pass’s record of it at 04:34 EDT, and it said the control was gone.
What the purge took
Section titled “What the purge took”The scratchpad is emptied by a nightly purge. The three helpers had been purged, silently, at some midnight before the audit.
A caution box went onto the top of the memory file. It quoted the superseded text, said the helpers went with the nightly purge, and then said what the rule had become: “the rule is back to a habit.” Until a helper in the repository did the job, the rule was to run the command that prints the time and copy what it printed.
Rebuilt where the purge cannot reach
Section titled “Rebuilt where the purge cannot reach”Joe authorised rebuilding it. His word was “rebuild it.”
Bilby rebuilt it inside the repository, where the nightly purge has no reach. The new posting tool fills the clock placeholder at post time, refuses a post that still has a leftover placeholder in it, and warns when someone has typed a clock time by hand. When it prepends a box to an existing post it keeps the original body underneath, and it refuses a retry that would stack the box twice. The tool that posts an approving review, the review tool, reuses the same fill function, loaded and not copied.
In the commit message Bilby put the loss briefly: the helpers “did this until” something deleted them, and “the memory kept calling them the control.” The rebuild’s own checks came to 13 of 13 for the posting tool and 22 of 22 for the review tool. The work was committed at 04:54:19 EDT.
Reviewing the diff
Section titled “Reviewing the diff”Skippy reviewed the diff and did not take the commit message on trust. He ran both self-tests, the 13 checks and the 22. Then he fetched the posting tool’s first live comment directly and compared its stamp with the server’s own creation time. The comment was stamped 04:53 EDT. The server had it created at 04:53:31-04:00. The two matched to the minute.
He also read the substitution logic and the retry guard line by line, because “selftest passes” was not going to be the whole story. Clean build, no findings.
stamp 04:53 EDT vs created_at 04:53:31-04:00